Skip to content

Welcome to ExE

ExE logo dark mode ExE logo light mode

What is the Experiments Engine?

Experiments Engine (ExE) enables social sector orgs with digital offerings to create and monitor experiments.

Features

  • Intuitive UI to create and monitor experiments
  • Dashboard to monitor all your experiments in one place
  • Get notified when events (e.g. number of trials run, number of days passed, etc.) occur
  • Track users across experiments and arms
  • Multiple methods: A/B testing, multi-armed bandit, and more

Want other features not listed here? Raise an issue in our GitHub repository or reach out to us at dsem@idinsight.org

Why ExE?

As penetration of smart phones and internet has increased in the Global South, there has been a surge in the number of digital offerings by social sector organizations.

Platform owners have numerous ideas on how to improve user experience and increase engagement but are hampered by the lack of technical expertise and resources to test these ideas.

We are building ExE to reduce these barriers to experimentation. We want to make it easy for platform owners to test out their ideas and make data-driven decisions. Our goal is to help orgs iterate, learn, and improve their offerings at low cost.

What is it not?

To measure the impact of your offering as a whole, you may need to do an impact evaluation. This tool does not replace that. Instead, it allows your to tweak nodes in your theory of change.

If you are looking for unbiased estimates of average treatment effects (say you want to publish an academic paper) this tool is not for you. Instead, this tool is for platform owners who want to quickly learn what works and what doesn't.

Who is it for?

ExE is for social sector orgs - NGOs, social enterprises, and developing country governments - who have digital offerings and a keen desire to improve them.

What kind of ideas can I test out?

Any idea where can both implement and measure the outcome digitally can be tested out using ExE.

Here are some ideas our partners are interested in testing out:

  • Does the personalised AI response lead to better user engagement?
  • Does using a casual tone with younger users lead to higher completion rates?
  • Does reducing the number of questions asked during onboarding lead to lower drop-off rates?
  • What is the best time to send messages to ensure users engage with it?

  • Will this new feature lead to greater time spent on the platform?

    Design principles

Here are the principles that guide the design of ExE:

  • Simple: You don't need to be statistician or engineer to use ExE. It is designed for non-technical platform owners to create and monitor experiments.
  • Light-weight: ExE is designed to be light-weight. We want to make it cheap for organizations to use.
  • Minimal data collection: ExE should store minimal data about users.
  • Open-source: ExE is open-source. We want to make it easy for organizations to use and contribute to the tool.

Who are we?

ExE is built by a team of data scientists, engineers, and statisticians at IDinsight. IDinsight's mission is to amplify the impact of our partners by using data and evidence to inform decisions.